Hemory AI Review: #1 Ambient Memory App Tested
Hemory Review: Always-On Memory for AI Agents
Key Takeaways
- Hemory, built by ONES.com founder Yingqi Wang, hit #1 Product of the Day on Product Hunt with 211 upvotes.
- It replaces meeting transcription with always-on capture on iPhone, Android, or Apple Watch, no extra hardware needed.
- Conversations auto-split into speaker-labeled moments, then feed Claude, Codex, Cursor, and other agents over MCP.
- Plans run from a free 10-hour trial to $50 a month for unlimited listening, billed only for detected speech.
- The company claims zero cloud audio storage. At barely a week old, independent reviews and a desktop release are still pending.
What Is Hemory, and Why Did It Hit #1 on Product Hunt?
A phone that listens to every conversation you have sounds like a battery drain and a privacy risk rolled into one product. Any honest Hemory AI memory review has to test that claim before it tests anything else, and the app's Product Hunt debut suggests plenty of people wanted to see the receipts too.
Hemory launched on Product Hunt, took the #1 Day Rank with 334 upvote points, and had already picked up 668 followers inside its first day. That is a rare showing for a brand-new consumer app in a crowded "AI memory" category that already includes Limitless, Granola, and half a dozen note-taking tools chasing the same idea.
The founder is not a stranger to enterprise software. Yingqi Wang is also the CEO of ONES.com, a project and R&D management platform founded in 2015 that has raised close to $100 million across three funding rounds within six months, including a $50 million Series C led by GIC. Hemory grew out of a small team inside that company, not a standalone startup chasing first product-market fit.
By his own account on the launch thread, Wang built Hemory because he spends half to two-thirds of his workday in meetings and kept forgetting to hit record, or left it running and ended up with several conversations merged into one giant transcript. The team spent nine months on the fix.
How Does Hemory's Always-On Listening Actually Work?
Hemory runs in two modes: Manual, where a person turns capture on only when they want it, and Schedule, where it listens automatically during set hours, such as weekdays from 9 a.m. to 6 p.m. Once it is listening, the day auto-splits into separate moments with speaker labels attached in real time, rather than one long undifferentiated recording.
The app separates conversations by activity as they happen, so a morning meeting, a lunch, and an afternoon walk each become their own searchable entry instead of one merged file. It runs today on iPhone, Android, and Apple Watch, all through hardware a person already owns. macOS, Windows, Linux, a web app, and a self-hosted option are all listed as coming soon, not shipped yet.
Where Does the Audio Actually Go?
Hemory's privacy page makes a specific claim, not a vague one. The company states audio is never stored in the cloud, that it is processed as a stream and destroyed the moment processing ends, and that any raw audio saved lives only on the device that recorded it, never synced across devices.
The founder added on Product Hunt that no audio files are stored by default, that listening can be switched off with one tap, and that anything already captured can be deleted at any time.
That is a strong claim for an ambient listening product, and it deserves to be treated as exactly that. A claim. Nothing in Hemory's public materials points to an independent security audit yet, so anyone recording client calls or anything covered by confidentiality rules should read the actual privacy policy before running Schedule mode for a full workday.
How Much Battery Does All-Day Listening Use?
Asked directly on the launch thread, Wang said 12 hours of listening on an iPhone 17 used about 20 percent more battery than normal, with no overheating on either the phone or the watch, and that an Apple Watch Ultra 3 still had more than half its battery left after 12 hours with other notifications turned off.
He also flagged a real limitation: a phone call or video meeting can take over the iPhone's microphone and interrupt capture, which is exactly where the Apple Watch's independent microphone keeps listening instead. That is a meaningfully different failure mode than a dedicated recording pendant, and it is the kind of detail most buyers only learn after the fact.
How Does Hemory Save Time for Remote Teams and Founders?
The use cases lean hard into founder and small-team workflows rather than enterprise meeting rooms. Wang's own examples were asking an agent what interesting things it heard yesterday, generating a half-year performance review from months of conversations, and pulling together every pricing discussion from the past month. Hemory's own marketing repeats that pattern: a monthly report deck built from a month of overheard conversations, a nightly journal entry generated automatically at 10 p.m., and a product requirements document assembled from a batch of customer interviews.
For a founder juggling calls, hallway conversations, and half-formed ideas said out loud on a walk, that turns Hemory into a continuous listening AI app for founders rather than a meeting-specific tool. A remote team spread across time zones gets a version of the same benefit. Instead of chasing down who said what on a call three weeks ago, someone asks an agent to search the memory directly.
Whether ambient memory tools like this one actually save founders hundreds of hours, as the pitch often goes, has not been independently measured for Hemory specifically. Treat it as a plausible mechanism, not a verified number.
How Does Hemory Connect to Claude, Codex, and Other AI Agents?
This is the feature that separates Hemory from a transcription app with a chatbot bolted on. Hemory connects to Claude Code, Codex, Gemini, Cursor, VS Code, OpenClaw, Hermes, and any other standard MCP client, using the Model Context Protocol, the open standard for linking AI tools to outside data sources.
Setup shown in Hemory's own documentation is a single command run once inside a supported agent, after which the agent gains a memory-search tool it can call on its own. In practice, a developer can ask Codex or Claude Code to pull the exact pricing conversation from two weeks ago instead of scrolling a transcript. The agent does the searching itself.
That design also explains why Hemory positions itself as a searchable ambient memory agent for meetings, not a meeting notetaker. A notetaker hands a human a document. Hemory hands an agent a queryable memory, a different product built for a different buyer, closer to the automation and no-code stacks this publication usually covers than to a meeting-room audience.
How Does Hemory Compare to Otter.ai, Apple Voice Memos, and Manual Note-Taking?
| Criteria | Hemory | Otter.ai | Apple Voice Memos | Manual note-taking |
|---|---|---|---|---|
| All-day, continuous capture | Yes, up to 24 hrs/day on Max, auto-split into moments | No, joins scheduled Zoom, Meet, or Teams calls, or records manually | No, manual recording only, no scheduling | No, limited to what a person types in real time |
| Audio retention | Company states zero cloud storage, raw audio stays on-device | Cloud-based recording and transcript storage under the account plan | Local by default, iCloud sync if enabled | Fully local, whatever app the person already uses |
| Search past conversations | MCP-connected agents can query weeks of moments by topic or speaker | Yes, inside Otter's own transcript archive | Limited, on-device title and text search | None, unless manually indexed |
| Hardware dependency | iPhone, Android phone, or Apple Watch already owned | Any device with the app, plus a calendar and video-call account | Apple devices only | Pen, notebook, or any notes app |
| Entry pricing | Free 10-hr one-time trial, then $10 to $50 per month | Free 300 min/month, Pro from about $8.33/mo billed annually | Free, bundled with the device | Free, cost is time |
| Built for AI-agent workflows | Yes, native MCP link to Claude, Codex, and Cursor is the core pitch | No native agent-memory layer | No | No |
Otter's free Basic plan covers 300 transcription minutes a month with a 30-minute cap per conversation, its Pro plan starts around $8.33 a month billed annually with 1,200 minutes, and its Business plan runs roughly $19.99 to $30 per user a month for unlimited meeting capture.The architectural difference is that Otter's calendar-linked bot joins scheduled Zoom, Google Meet, or Microsoft Teams calls, built around virtual meetings by design. Hemory listens to whatever is happening near the device, meeting or not. Manual note-taking still wins on zero cost and zero setup. It loses on everything else a search bar can fix, including the fact that human memory of a conversation degrades within hours.
What Are Hemory's Biggest Limitations Right Now?
None of this makes Hemory a finished product yet. At the time of writing it carries a single Product Hunt review, not enough data to judge real-world reliability. Android users install it from a direct APK download rather than the Google Play Store, and desktop apps, a web client, and self-hosting are all still on the roadmap rather than shipped.
Hemory has not published an independent search-accuracy or transcription-accuracy benchmark yet, so claims about finding the right moment quickly still rest on company demos rather than third-party testing. The VAD billing model is also worth reading closely: it protects a quiet solo user's quota, but a back-to-back day of group meetings still counts as detected speech for most of its duration, so heavy multi-person use will burn hours faster than the marketing framing suggests. And any product recording other people's voices, coworkers, clients, family, puts a consent question on the user, not the app.
Frequently Asked Questions:
Q. What is Hemory and how does it work?
Hemory is an always-on listening app for iPhone, Android, and Apple Watch that turns everyday conversations into searchable, speaker-labeled memory, then connects that memory to AI agents such as Claude and Codex over MCP.
Q. Is Hemory always recording, and what happens to the audio?
It listens continuously only in Schedule mode, or on demand in Manual mode. The company states audio is processed as a stream, never stored in the cloud, and kept only on the recording device until the user deletes it.
Q. How much does Hemory cost, and what counts against my usage quota?
Plans range from a free one-time 10-hour trial to a $50-a-month Max plan with unlimited listening. Billing uses voice-activity detection, so silence and background noise never count against the quota, only detected speech does.
Q. How much battery does continuous listening use on iPhone and Apple Watch?
The founder reported roughly 20 percent extra battery drain over 12 hours on an iPhone 17, with no overheating, and more than half a charge remaining on an Apple Watch Ultra 3 after 12 hours with notifications off.
Q. How does Hemory connect to Claude, Codex, or other AI agents?
Through the Model Context Protocol. A single setup command inside a compatible agent adds a memory-search tool it can call directly, letting it pull specific past conversations instead of a human searching a transcript.
Disclaimer: This article reflects publicly available information about Hemory as of late September 2026 and is not investment, legal, or purchasing advice. Pricing, features, and privacy practices are set by Hemory and ONES.com and may change without notice. The Dollar Craft received no compensation from Hemory, ONES.com, or any competitor named in this review.
Comments
Post a Comment